The most obvious thing about languages is how different they all seem to be – as is especially obvious to anyone trying to learn a new language as an adult. Reflecting a general consensus on this point in 1957, Martin Joos propounded that “languages can differ from each other without limit and in unpredictable ways” (In Hamp et al (1995)). It might seem that it is this apparently limitless variability in language which gives some children the problems they evidently have. But also in 1957 Noam Chomsky made an exactly opposite proposal, intrinsically stronger because it is falsifiable. So it is the weaker claim by Joos which needs defending now, if indeed it can be defended against the now substantial evidence that there are significant language universals.

The biolinguistics assumed here directly descends from Chomsky’s 1957 proposal. From this perspective, as the 1957 proposal has been successively modified in the light of different sorts of evidence, it now makes sense to speak of what Noam Chomsky (1995) calls “Computation for Human Language’.

Even in severe disorder, like speech disorders in which the speaker is almost entirely incomprehensible even to the speaker’s parents, the hierarchy still holds. Obviously this has to be scrutinised, case by case. But with a narrowly defined hierarchy of branching, scrutiny is possible, at least in principle.

But by the proposal here, there was never any inevitability. Evolution happens by small steps which happen to contribute to the organism’s survival to the point of reproduction. And in the case of speech and language, these small steps took the form they did because they were formatted in such a way that they could be encoded by the genome. The evolutionary history could have been quite different. As artificial languages have been developed for particular purposes, in all cases very narrowly defined, to express principles of logic, to facilitate financial accounts, to teach basic programming, they have taken quite different characters. Given that there is evidence that some cetaceans identify particular members of the pod by uniquely identifying call signs, it would be unsurprising if the same thing had happened in human speech and language evolution. Plainly, this referential property was and is important. All human languages have nouns. But this property of referentiality, expressed in the category of nouns, could have become the central defining property of grammatical structure. By such a grammar, all words could vary only in their degree of referentiality, either in two ways or in three or more ways. The simple command “Go” would be either totally non-referential or uninterpretable. Human language happened not to follow this particular direction. But it could have done.

A language with a grammar defined exclusively on degrees of referentiality would fall outside the terms of the proposal here. It would not be a possible human language. But it would seem possible to me in principle that a child might consider such an analysis of his or her target language. Such an analysis would not get the child very far. It would not parse even the simplest sentence. But a child could make this profound, though very unlikely, mistake.

Or the child could make any correspondingly profound mistake such as trying to analyse the whole of language on the basis of one of its components, such as topicalisation, as by “Snails, I absolutely refuse to even try” where snails is a topic and the rest of the structure is a comment, or the marking of respect on the one hand and familiarity on the other in accordingly different pronouns for you, or the definition of time-scale or evidentiality, or any of the other semantically and pragmatically crucial properties expressed in various ways in the grammar of English and many other languages. But none of these properties are universal. Rather they are part of an apparatus by which they can be expressed in a number of language-specific ways.

With any child. with any degree of impairment in speech or language, it is a matter of investigation whether the impairment falls outside the terms of the universal hierarchy. For some linguists this may seem like an impossibility. But it seems to me that it should not be disallowed in principle, no matter how rare and unlikely it may happen to be.